Dentomaxillofacial Radiology (2007) 36, 1217 q 2007 The British Institute of Radiology http:/ /dmfr.birjournals.

org

RESEARCH

Comparison of diagnostic accuracy of lm and digital tomograms for assessment of morphological changes in the TMJ
M Wiese*,1,2, H Hintze1, P Svensson3 and A Wenzel1
1 Department of Oral Radiology, University of Aarhus, Aarhus, Denmark; 2Department of Radiology, University of Copenhagen, Copenhagen, Denmark; 3Department of Clinical Oral Physiology, University of Aarhus, Aarhus, Denmark

Objective: To compare diagnostic accuracy of tomograms obtained with lm and three digital image receptor modalities for detection of morphological changes in the temporomandibular joint (TMJ). Methods: Lateral and frontal cross-sectional tomograms of 158 TMJs in 80 dry human skulls were obtained with four X-ray receptors: one conventional lm (Agfa-Gevaert), two photostimulable phosphor (PSP) plates: VistaScan and Digora PCT and one charge-coupled device (CCD): ProMax. The lm and the PSP plate tomograms were exposed in a Cranex Tome X-ray unit and the CCD tomograms in the ProMax X-ray unit. The tomograms were examined by three observers for the presence of morphological changes in the condyle, the mandibular fossa and the articular tubercle. Naked-eye inspection of the articular surfaces of the TMJs performed by the same three observers served as the gold standard for the true presence of morphological changes. Results: For examination of TMJ changes using lateral and frontal tomograms in combination and lateral tomograms separately, the ProMax modality obtained a signicant lower diagnostic accuracy than the other three modalities, between which no signicant differences were found. For morphological changes localized to the condyle in which the highest number of changes were present according to the gold standard, the same result was found. Conclusion: The ProMax modality was signicantly less accurate than the lm, VistaScan and Digora PCT modalities for the detection of overall TMJ morphological bone changes as well as condylar bone changes. Dentomaxillofacial Radiology (2007) 36, 1217. doi: 10.1259/dmfr/78486936 Keywords: radiography; temporomandibular joint; digital radiography; modalities; tomography Introduction Radiographic examination of the temporomandibular joint (TMJ) is often included in the diagnostic procedure of patients suffering from temporomandibular disorders.1,2 Among the most widely used conventional imaging techniques, cross-sectional tomography is known to be the technique revealing the greatest number of structural changes and representing most precisely the anatomic structures of the TMJ.1 The capacity of the X-ray unit may determine the diagnostic outcome of the tomographic examination, i.e. complex X-ray tube motions (spiral and hypocycloid instead of linear) and individual angle settings (instead of standard settings) have shown to be important
*Correspondence to: Mie Wiese, Department of Radiology, School of Dentistry, 20, DK-2200 Faculty of Health Sciences, University of Copenhagen, Noerre Alle Copenhagen N, Denmark; E-mail: mwi@odont.ku.dk Received 28 September 2005; revised 3 February 2006; accepted 1 March 2006

for a good image quality and hence increase the probability of an accurate diagnostic outcome.1,3 Whether the image receptor used for TMJ tomography has an inuence on the diagnostic outcome, too, does not seem to have been evaluated. Over the past two decades, digital image receptors have been developed for replacement of the conventional lm. Currently, there are two well-established digital modalities available on the market: the charge-coupled device (CCD) and the photostimulable phosphor (PSP) plate, the latter to be used with a conventional radiographic unit for extraoral examination, whereas the former usually is part of a dedicated digital radiographic unit. While several studies have compared the diagnostic outcome of those two fundamentally different digital receptors for intraoral examinations,4,5 only a few studies have focused on their effectiveness when used for extraoral examinations. In a

TMJ tomography M Wiese et al

13

study by Benediktsdottir et al,6 no signicant difference in diagnostic accuracy was found between digital panoramic images, obtained with two CCD and PSP plate modalities, and conventional lm for the assessment of mandibular third molars. Whether the accuracy differed among the various digital modalities was not tested, but large differences between their image qualities were demonstrated.7 Apart from those two studies, however, little is known about the inuence of digital image receptors on the diagnostic outcome of extraoral examinations used for various diagnostic tasks. Therefore, the aim of this study was to compare the diagnostic accuracy of tomograms obtained with lm and three digital image receptor modalities for detection of morphological changes in the TMJ.

Materials and methods The left and right TMJ in 80 dry skulls were examined radiographically by lateral and frontal tomography with one conventional lm: Curix Ortho HT-G (Agfa-Gevaert NV, Mortsel, Belgium), and three digital modalities; two rr Dental, Bietigheim-Bissingen, PSP plates: VistaScan (Du Germany) and Digora PCT (Soredex, Helsinki, Finland), and one CCD: ProMax (Planmeca OY, Helsinki, Finland). Details regarding exposure parameters etc. for the different receptors are presented in Table 1. The lm and the PSP plate tomograms were exposed in a Cranex Tome radiographic unit (Soredex, Helsinki, Finland) and the CCD-based tomograms were exposed in the ProMax radiographic unit (Planmeca OY). The Cranex Tome unit makes use of corrected TMJ tomography (adapted to the individual joint) with the tomographic plane perpendicular to the long axis of the condyle for lateral tomograms and the tomographic plane parallel to the long axis of the condyle for frontal tomograms. The angle of the tomographic plane is decided from a lateral four-angle preexamination. With the ProMax unit, it is nearly impossible to perform corrected lateral and frontal TMJ tomography (for explanation see Discussion), so instead the tomograms were obtained by visual assessment of the
Table 1 Exposure procedures and parameters used for the four modalities Layer thickness (mm), slide distance (mm), number of cuts Modality Film VistaScan X-ray unit (tube motion) Cranex Tome (spiral) Cranex Tome (spiral) Lateral tomograms 4, 3, 4 4, 3, 4

localization of the laser light beams, indicating the tomographic plane in relation to the long axis of the condyle. In both radiograph units, the skulls were xed in a photostat device to ensure optimal positioning for the equipment. The lateral tomograms were performed with the mandible in a stable dental occlusion. The frontal tomograms were performed with the mandible protruded and the mandibular incisors slightly anterior to the maxillary incisors. The condyle was positioned below the articular tubercle, aiming at keeping the superior surface of the condyle free of superimposition of the articular tubercle. For skulls with tooth loss, a stable occlusion was established using a silicone impression. The examinations with the Cranex Tome unit were performed by a radiographer who had experience with working with this unit, whereas examinations with the ProMax unit were performed by a dentist instructed in the use of this unit. Examples of corrected lateral and frontal images using the different receptors are shown in Figure 1. After a calibration session, three observers examined the lateral and frontal tomograms for the presence of attening and erosive changes in the condyle and articular tubercle using the following denitions: attening, loss of convexity or concavity of the joint outlines; erosion/defect, a local area of rarefaction in the layer of compact bone.8 Osteophytes (local outgrowth of bone arising from the mineralized surface8) in the condyle and changes (attening and erosion) in the mandibular fossa were assessed in lateral tomograms only. Multiple examples of attening, erosion and osteophyte in TMJ tomograms were collected in an atlas, which the observers could consult in case they were in doubt how manifest changes should be to be scored. The lm tomograms were examined on a light box using a radiograph viewer with magnication, while the digital tomograms were examined in their own software on an IBM 19 inch cathode-ray tube monitor placed in a room with subdued lighting. There were no restrictions as to which enhancement facilities in the software the observers could use. Two experienced oral radiologists (Observers 1 and 2) and one general dentist with little experience with radiography (Observer 3) independently examined the lateral and frontal tomograms of each joint with each

Frontal Magnication Exposure tomograms factor settings 6, 5, 4 6, 5, 4 1.5 1.5

Radiation dose lter

Screens

Development Kodak X-OMAT 3000 RA Vista scanner

Digora PCT Cranex Tome 4, 3, 4 (spiral) ProMax ProMax (linear) 6, 9, 3

6, 5, 4 6, 9, 4

1.5 1.5

57 66 kV, Four layers Lanex Regular 1.6 5.0 mA of plastic foil (Eastman Kodak Co., Rochester, NY) 57 66 kV, Four layers None (but one 1.6 5.0 mA of plastic foil non-specic PSP plate in front of VistaScan plate) 57 66 kV, Four layers None 1.6 5.0 mA of plastic foil 57 kV, None None (sensor) 3.0 4.0 mA

Digora PCT scanner None (direct)

Dentomaxillofacial Radiology

14

TMJ tomography M Wiese et al

Figure 1

Examples of corrected lateral and frontal tomography using lm and three digital image receptor modalities

image modality. The left and right joints in each skull were examined independently of each other. A binary registration scale where Score 0 indicated not certain of change and Score 1 indicated certain of change was used. In addition, the image quality of the tomograms was assessed as being either sufcient or insufcient. A month after the radiographic evaluation, the three observers re-evaluated 10% of the tomograms for the assessment of intraobserver agreement. Three months prior to the radiographic examinations, all TMJs had been validated for the presence of morphological TMJ changes by the same three observers. A calibration session preceded the individual observer examination. The surfaces of the condyle, mandibular fossa and articular tubercle were examined by naked-eye inspection for macroscopic changes (erosion) and deviations in form (attening and osteophyte) using a magnifying viewer with built-in light. Score 0 indicated not certain of change and Score 1 indicated certain of change. A change was dened as present if reported by at least two of the three observers. The naked-eye inspection served as gold standard for the radiographic examinations. Data treatment First, recordings from the lateral and frontal tomograms were evaluated separately to calculate the accuracy for
Dentomaxillofacial Radiology

each projection. Second, recordings from the lateral and frontal tomograms were assessed in combination to evaluate the accuracy of an examination including both projections. For the combined examination, changes which could appear in both projections (attening and erosive changes in the condyle and the articular tubercle) were dened to be present if they were recorded just from one of the projections, whereas changes appearing in one projection were included only in case they were recorded from the relevant projection (osteophyte in the condyle, and attening and erosive changes in the mandibular fossa were recorded from lateral tomograms only). Perfect diagnostic accuracy was dened as agreement between the gold standard and the radiographic assessment resulting in a score difference of 0 (subtracting the radiographic score from the gold standard score). Disagreement between the gold standard score and the radiographic score resulted in a score difference of 1 or 2 1, where 1 expressed a radiographic false-negative recording and 2 1 a radiographic false-positive recording. The diagnostic accuracy of the image modalities was expressed as the mean cumulative score calculated for each of the three observers. Thus, the lower the mean cumulative score, the higher the diagnostic accuracy. Calculations were based on an overall diagnosis of the joint and on the different joint components separately, i.e. the pooled scores of all

TMJ tomography M Wiese et al

15

Table 2 Frequency of TMJs with morphological changes according to the gold standard Frequency Morphological change Condyle Flattening Erosion/defect Osteophyte Mandibular fossa Flattening Erosion/defect Articular tubercle Flattening Erosion/defect % 40 59 8 0.6 4 4 15 N 63 93 13 1 7 6 23

Results At the end of the radiographic exposures, two joints missed examination with all four image receptors. These two joints were excluded from the study material, which thereafter included 158 joints in 80 skulls. The number of condyles, mandibular fossae and articular tubercles with morphological changes according to the gold standard are shown in Table 2. Figure 2 illustrates the accuracy of the four modalities expressed as the mean cumulative score of the three observers for the overall detection of TMJ changes. Evaluating the lateral and frontal tomograms in combination, ProMax had a signicantly lower accuracy than the other three modalities, between which no signicant difference was found. The same result was obtained for lateral tomograms evaluated separately. Using frontal tomograms only, ProMax was signicantly less accurate than lm and VistaScan (P values are listed in Figure 2). Between the PSP plate modalities and the lm, no signicant differences were found. The observers individual performances with the four modalities are presented in Table 3. In general, Observer 1 had a higher cumulative score and thereby a lower accuracy than Observers 2 and 3, Observer 3 having the lowest score (the highest accuracy). Overall, this was independent of the image modality and whether the lateral and frontal tomograms were evaluated in combination or separately. The difference in cumulative scores for the three observers was signicant in the lateral and frontal tomograms assessed in combination (P , 0.001) and the frontal tomograms assessed separately (P 0.001). The accuracy of the four modalities in detecting morphological changes localized to the condyle, the mandibular fossa and the articular tubercle, respectively, from lateral and frontal tomograms in combination is presented in Figure 3. For the detection of bone changes in the condyle, ProMax had a signicantly higher cumulative score and thereby a lower accuracy than the remaining modalities, between which no signicant differences were found. For detection of bone changes in the mandibular fossa and the articular tubercle, no signicant difference in accuracy between the modalities was found. Mean kappa for intraobserver agreement for morphological changes localized to the condyle on lateral tomograms was: 0.82 for Observer 1 (range 0.43 1.00), 0.87 for Observer 2 (range 0.63 1.00) and 0.98 for Observer 3 (range 0.76 1.00). The corresponding values from frontal tomograms were: 0.75 for Observer 1 (range 0.43 1.00), 0.65 for Observer 2 (range 0.29 1.00) and 0.83 for Observer 3 (range 0.66 1.00). Kappa was 1.00 for all observers for the assessment of the mandibular fossa and articular tubercle. Film and ProMax examinations were recorded to be of insufcient quality in 8.2% of the cases, followed by the Digora PCT with 4.4% and the VistaScan with 1.9% insufcient tomographic examinations. The frequency of insufcient examinations with lm and ProMax was signicantly higher than the frequency reported with VistaScan (P , 0.02), but not signicantly different from
Dentomaxillofacial Radiology

individual ndings of attening, osteophyte and erosive changes. Differences between the receptor modalities were tested by two-way analysis of variance (ANOVA) with observer and modality as the two independent variables. Pairwise comparisons were made by post hoc t-test (SPSS package, GLM, version 10.0 for Windows, SPSS Inc., Chicago, IL). Intraobserver agreement was expressed as kappa values. For the assessment of image quality, an entire joint examination was dened as consisting of a lateral and a frontal tomogram, and if the quality of either projection was reported as insufcient, the quality of the whole examination was dened as being insufcient. Insufcient image quality was dened to be present if reported by two of the three observers. Differences in number of examinations with insufcient image quality were tested by x 2 test. The level of signicance was set at P , 0.05.

Figure 2 Accuracy of four modalities for detection of morphological changes in the TMJ expressed as the mean cumulative score of three observers using different tomographic projections (the lower the cumulative score, the higher the diagnostic accuracy)

16
Table 3

TMJ tomography M Wiese et al

Individual observer performance with four modalities expressed as the cumulative score (the lower the score, the higher the diagnostic accuracy) Lateral and frontal tomograms in combination, Observer Lateral tomograms separately, Observer 1 195 192 193 205 2 194 178 189 199 3 182 184 189 195 Frontal tomograms separately, Observer 1 161 166 168 177 2 167 156 163 171 3 130 129 144 154

Modality Film VistaScan Digora PCT ProMax

1 188 186 190 206

2 187 169 178 197

3 143 143 159 173

the frequency reported with Digora PCT. Between the Digora PCT and VistaScan, no signicant difference was present. Discussion For radiographic imaging of the TMJ, tomography is considered the most accurate conventional technique.9 12 Traditionally, TMJ tomography is performed only in the sagittal plane as lateral tomograms, although several authors have documented the importance of also including coronal views in the form of frontal tomograms oriented perpendicularly to the lateral projection.10,13 16 In modern panoramic radiographic units, programmes for both lateral and frontal TMJ tomography might be available. The technical procedures and the practical steps associated with these programmes in the various radiographic units vary signicantly, however. In order to orient the tomographic plane parallel or perpendicular to the long axis of the condyle, some units have a programme for a rough assessment of the deviation of the condyles long axis, which subsequently supports the selection of programmes giving the correct orientation of the tomography plane; others have no such facility. The Cranex Tome radiographic unit has a programme to determine the individual condyle deviation in relation to the horizontal plane, a programme which is very easy for the radiographer to operate. The ProMax unit has a corresponding programme showing the condyles deviation in relation to the frontal plane (frontal three-angle pre-examination), but in practice it is very difcult to convert the parameters from this pre-examination to the subsequent tomographic examinations. Due to this difculty, the pre-examination with the ProMax unit was omitted in the present study, and the lateral and frontal tomographic examinations were performed directly, selecting the wanted tomographic planes on the basis of the units aiming laser light beams. Expressing the accuracy of the different modalities as the number of radiographic scores differing from the gold standard resulted in a single-valued parameter, which provided a greater separation of the modalities than using traditional sensitivity and specicity values. With this denition of accuracy, the present results showed a signicantly lower diagnostic accuracy with the ProMax modality using lateral and frontal tomograms in combination and lateral tomograms separately for an entire TMJ examination than with lm, VistaScan and Digora PCT. Using frontal tomograms separately, ProMax was signicantly less accurate than lm and VistaScan. Independent
Dentomaxillofacial Radiology

of the projections, the highest accuracy was obtained with VistaScan, for which the mean cumulative score was in general 2 6% lower than the corresponding scores for lm and Digora PCT. The difference was, however, not signicant. It may be surprising that the lm tomograms were as accurate as the PSP modalities, since no image enhancement facilities can be applied during the examination. When the observers examined the digital tomograms, magnication and density, and contrast enhancement for the various joint components were used constantly. However, this did not increase the diagnostic accuracy. On lateral and frontal tomograms in combination, the individual observer obtained a cumulative score with the ProMax modality being 5 9% higher and thereby less accurate than the modality with the second highest score. This tendency was present also when lateral and frontal tomograms were evaluated separately. It seemed especially to be a problem to make a correct detection with ProMax in the condyles, for which the mean cumulative score was 12% higher than the scores found with lm and Digora PCT, and 18% higher than the score found with VistaScan. In the mandibular fossa and the articular tubercle, the differences between the modalities were insignicant as

Figure 3 Accuracy of four modalities for detection of morphological changes in the various TMJ components expressed as the mean cumulative score of three observers using lateral and frontal tomograms in combination (the lower the cumulative score the higher the diagnostic accuracy)

TMJ tomography M Wiese et al

17

expected. Since those joint components had very few morphological changes, a general negative radiographic recording would result in a subtraction score (gold standard score radiographic score) of 0. From previous studies, it has been found more prevalent to underestimate than to overestimate the presence and extent of TMJ changes.10,11,17,18 This fact seemed to be valid also in this study, in particular assessing the mandibular fossa and articular tubercle. The three observers performances with the different modalities varied signicantly. In particular, it was surprising that Observer 3a general dentist with little radiographic experiencedetected bone changes more correctly with all modalities than the two other observers, being oral radiologists with several years of experience. Usually, the most accurate diagnoses are expected from observers with great experience, but the present results did not reach this expectation. The reason why Observer 3 performed more accurately than the remaining observers is, however, unknown. The mean intraobserver agreement was almost perfect for all the three observers when they used lateral tomograms for the assessment of morphological changes, and it was substantial when frontal tomograms were used.19 Tomograms of insufcient quality were recorded signicantly more often with lm and ProMax than with VistaScan. The main reason for the insufcient quality with the lm modality seemed to be density problems. In some tomograms, the observers found the density too high to allow a simultaneous examination of the condyle, the
References
, 1. Brooks SL, Brand JW, Gibbs SJ, Hollender L, Lurie AG, Omnell K-A Westesson P-L, White SC. Imaging of the temporomandibular joint. A position paper of the American Academy of Oral and Maxillofacial Radiology. Oral Surg Oral Med Oral Pathol Oral Radiol Endod 1997; 83: 609618. 2. Pharoah MJ. The prescription of diagnostic images for temporomandibular joint disorders. J Orofac Pain 1999; 13: 251 254. 3. Westesson P-L, Katzberg RW. Imaging of the temporomandibular joint. Baltimore: Williams & Wilkins; 1991. Vol. 1, Number 1. 4. Hintze, H. Radiography for the detection of dental caries lesions. Thesis. University of Aarhus 2004 5. Wenzel A. Current trends in radiographic caries imaging. Oral Surg Oral Med Oral Pathol Oral Radiol Endod 1995; 80: 527 539. ttir IS, Hintze H, Petersen JK, Wenzel A. Accuracy of 6. Benediktsdo digital and lm panoramic radiographs for assessment of position and morphology of mandibular third molars and prevalence of dental anomalies and pathologies. Dentomaxillofac Radiol 2003; 32: 109 115. ttir IS, Hintze H, Petersen JK, Wenzel A. Image quality 7. Benediktsdo of two solid-state and three photostimulable phosphor plate digital panoramic systems and their utility for treatment planning before mandibular third molar removal. Dentomaxillofac Radiol 2003; 32: 39 44. , Petersson A. Radiography of the temporomandibular 8. Omnell K-A joint utilizing oblique lateral transcranial projections. Comparison of information obtained with standardized technique and individualized technique. Odont Revy 1976; 26: 77 92. , O berg T. Comparison between radiologic 9. Bean LR, Omnell K-A observations and macroscopic tissue changes in temporomandibular joints. Dentomaxillofac Radiol 1977; 6: 90 106.

mandibular fossa and the articular tubercle. However, this perception did not inuence the accuracy of the lm, which did not differ signicantly from the PSP modalities. With the ProMax, the main reason for insufcient quality seemed to be the way that the joint appeared in the tomograms. In particular, the observers found it cumbersome that the magnication of the cuts in the same tomographic projection varied, and that the separation between the cuts was poorer than the other modalities. This is not directly due to the image receptor, but rather to the tomographic principle offered in the ProMax radiographic unit. Therefore, the higher number of insufcient examinations with the ProMax modality seemed to be caused in particular by the available technique rather than the image receptor. The very low number of VistaScan examinations having an insufcient quality may be due to the system software, which offers enhancement lters developed for particular diagnostic tasks. The lters are applied to the images very easily, and in the present study all three observers constantly used one of the pre-dened lters particular. In conclusion, the ProMax modality was signicantly less accurate than the lm, VistaScan and Digora PCT modalities for detection of overall TMJ morphological changes as well as condylar bone changes.
Acknowledgments We thank Senior Programmer Erik Gotfredsen, Department of Oral Radiology, University of Aarhus, for help with statistical analyses of the data.

10. Tyndall DA, Davies KL, Ludlow JB. A comparison of zonography, multidirectional tomography, and biplanar panoramic images for detection of simulated condylar arthropathy of the temporomandibular joint. Int J Adult Orthodon Orthognath Surg 1995; 10: 277 284. 11. Ludlow JB, Davies KL, Tyndall DA. Temporomandibular joint imaging: a comparative study of diagnostic accuracy for the detection of bone change with biplanar multidirectional tomography and panoramic images. Oral Surg Oral Med Oral Pathol Oral Radiol Endod 1995; 80: 735 743. kerman S. 12. Cholitgul W, Petersson A, Rohlin M, Tanimoto K, A Diagnostic outcome and observer performance in sagittal tomography of the temporomandibular joint. Dentomaxillofac Radiol 1990; 19: 16. 13. Rozencweig D. Three-dimensional tomographic study of the temporomandibular articulation. J Periodontol 1975; 46: 348 356. 14. Berrett A. Radiology of the temporomandibular joint. Dent Clin North Am 1983; 27: 527 540. 15. Sato H, Fujii T, Yamada N, Kitamori H. The contribution of frontal tomography to the diagnosis of temporomandibular joint osteoarthritis. Dentomaxillofac Radiol 1992; 21: 77 80. 16. Bouwman JPB, Kerstens HCJ, Tuinzing DB. Condylar resorption in orthognathic surgery. Oral Surg Oral Med Oral Pathol 1994; 78: 138 141. kerman S. Microscopy and tomography of 17. Flygare L, Rohlin M, A erosive changes in the temporomandibular joint. An autopsy study. Acta Odontol Scand 1995; 53: 297 303. kerman S, Kopp S. Tomography as an aid to detect 18. Rohlin M, A macroscopic changes of the temporomandibular joint. An autopsy study of the aged. Acta Odontol Scand 1986; 44: 131 140. 19. Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics 1977; 33: 159 174.

Dentomaxillofacial Radiology